<b>TgFC: </b><b><i>Tectona grandis</i></b><b> </b><b>Fungal Community Dataset</b>
收藏资源简介：
The <b>TgFC dataset</b> comprises <b>5,236 annotated microscopic images</b> (640×640 resolution) of fungal spores from diseased <i>Tectona grandis</i> (teak) leaves, collected in Bangladesh (Sylhet and Rangamati). The images represent three fungal taxa:<b><i>Olivea tectonae</i></b><b> (2,219 images)</b><b><i>Neopestalotiopsis</i></b><b> sp. (1,688 images)</b><b><i>Colletotrichum siamense</i></b><b> (1,329 images)</b>Spores were imaged at 40× magnification using a Zeiss Primostar 3 microscope. Wet mount preparations were made without staining to preserve field morphology. All images were manually annotated with bounding boxes using LabelImg for object detection, based on species-specific spore morphology.The dataset is structured for YOLO-based object detection and includes separate folders for training (80%), validation (10%), and testing (10%)—each containing <code>images/</code> and <code>labels/</code> subdirectories. No image preprocessing was performed, allowing flexible adaptation to custom pipelines.The dataset is annotated in YOLO format for object detection but can be adapted for various computer vision and image analysis tasks, including traditional machine learning, deep learning, classification, and segmentation workflows.<b>Code & Validation: </b><i>here</i><b>Applications and Impact</b><b>:</b>Enables automated spore identification from foliar samples, reducing manual diagnostic workloadSupports airborne spore detection and spore quantification in environmental samples for outbreak predictionIntegrates with real-time disease monitoring systems for early intervention in agricultureFacilitates training and benchmarking of deep learning models (e.g., YOLO, CNNs) for fungal detectionSupports transfer learning across fungal species and imaging conditionsProvides a foundation for developing portable, field-ready diagnostic toolsApplicable beyond teak, as the included fungi are common in other tropical and subtropical ecosystemsOpen-access format encourages future expansion and collaborative contributions of more annotated images<br><b>If you use this dataset or validation scripts, please cite:</b>The TgFC dataset paper is Under Review <i>(Journal: Scientific Data)</i>Cite the Dataset. DOI: 10.6084/m9.figshare.28855910<b>For questions or collaborations, please contact:</b>Maaz Ahmed – ahmedmaaz.maaz@gmail.comSyeda Munjiba Islam - islammunjiba@gmail.comMd. Faysal Ahamed - faysal.ahamed@ece.ruet.ac.bd
<b>TgFC数据集</b>包含<b>5236张带标注的显微图像</b>（分辨率为640×640），采集自孟加拉国锡尔赫特（Sylhet）和兰加马蒂（Rangamati）地区的患病柚木（<i>Tectona grandis</i>）叶片上的真菌孢子。该数据集涵盖3类真菌类群：<b><i>橄榄孢菌（Olivea tectonae）</i></b>（2219张）、<b><i>新拟盘多毛孢属（Neopestalotiopsis sp.）</i></b>（1688张）、<b><i>暹罗炭疽菌（Colletotrichum siamense）</i></b>（1329张）。 所有孢子均使用蔡司Primostar 3（Zeiss Primostar 3）显微镜以40倍放大倍率成像。制片采用湿封片法且未进行染色，以保留样本的野外原生形态。所有图像均通过LabelImg标注工具手动绘制边界框以用于目标检测任务，标注依据为各物种特有的孢子形态特征。 本数据集专为基于YOLO的目标检测任务设计，设有训练集（80%）、验证集（10%）与测试集（10%）三个独立文件夹，每个文件夹下均包含<code>images/</code>与<code>labels/</code>两个子目录。未对图像进行任何预处理，可灵活适配各类自定义图像处理流程。 本数据集采用YOLO格式进行标注，适配目标检测任务，同时可适配多种计算机视觉与图像分析任务，涵盖传统机器学习、深度学习、图像分类及图像分割等工作流。 <b>代码与验证：</b><i>此处</i> <b>应用与价值：</b> 可实现叶片样本中真菌孢子的自动化识别，减轻人工诊断工作量； 支持环境样本中的气传孢子检测与定量分析，用于病害暴发预测； 可集成于农业实时病害监测系统，实现病害早期干预； 可为真菌检测相关深度学习模型（如YOLO、卷积神经网络（Convolutional Neural Network））的训练与性能基准测试提供支撑； 支持跨真菌物种与成像条件的迁移学习； 为开发便携式田间诊断工具提供核心支撑； 本数据集涵盖的真菌类群广泛分布于其他热带与亚热带生态系统，因此其应用场景不限于柚木种植领域； 其开放获取的格式有助于未来扩充更多标注图像，推动社区协作贡献。 <b>引用说明：</b>若使用本数据集或验证脚本，请按以下方式引用： TgFC数据集论文目前正在《Scientific Data》期刊审稿中，请引用本数据集，DOI：10.6084/m9.figshare.28855910 <b>咨询与合作：</b>如有疑问或合作意向，请联系： 马扎·艾哈迈德（Maaz Ahmed）：ahmedmaaz.maaz@gmail.com 赛义达·蒙吉巴·伊斯兰（Syeda Munjiba Islam）：islammunjiba@gmail.com 穆罕默德·费萨尔·艾哈迈德（Md. Faysal Ahamed）：faysal.ahamed@ece.ruet.ac.bd
